Evaluation of the method for analysis of vicine, convicine and their aglycones

6.1.1 Extractability of vicine, convicine and their aglycones

Vicine and convicine were extracted from faba bean flours with two extraction solutions (7% PCA and water) and 1–3 repeated extractions in a total volume of 15 ml.

Deproteinisation was one criterion for extraction, because faba bean contains high amounts of proteins that are either soluble in water (albumins) or in salt solutions (globulins, as storage proteins). Extracts were deproteinised either with PCA or by adjusting the pH to 4, which is the known isoelectric point of most faba bean proteins (Bhatty 1974). The comparison showed that differences in extractability were small; however, certain differences were found. Repeated extraction (3 times) with 7% PCA was selected as the final method.

Slightly more vicine was measured in acidic extracts, whereas convicine was somewhat better extracted by water; however, the differences were small between the two extraction solutions. Marquardt et al. (1983) reported that vicine was more soluble at low pH and convicine at high pH values. The solubilities of these two compounds were lowest at pH values 4–8. Marquardt et al. (1983) also reported that 3–3.6 mg/ml of vicine could be solubilised at pH values 4–9 and 0.4–1.2 mg/ml of convicine at pH values 1–8. In the present faba bean analysis, the vicine and convicine concentrations were lower (0.1–0.2 mg/ml) in the extraction solution than their given solubilities; therefore, the differences in solubilities did not limit the analysis. The extraction can be carried out at pH values of 4–8, and acidic extraction is not necessary. Possible acid-induced degradation was not a limitation in the acidic extraction, because the results were repeatable and comparable between the two solutions.

Vicine and convicine were readily dissolved from faba bean flour, but repetition of the extraction as two or three cycles slightly increased the peak areas of the analytes. Repeated extraction particularly improved the extractability of convicine in acidic conditions, so the decision was made for a three-step extraction (3 × 5 ml) as the final method. The peak areas for uridine were also slightly increased by repeated extraction. A single extraction gave acceptable results, and would therefore be sufficient for faster screening of vicine and convicine levels.

The internal standard uridine was added to faba bean flours before the extraction solution, and its recovery was good. It was previously used as an internal standard by Quemener

62

(1988) and Helsper et al. (1993). The suitability of cytidine as an internal standard was estimated by Purves et al. (2018a), but cytidine was converted to uridine by cytidine deaminase in water solutions, indicating that enzyme inactivation would be required during the extraction. Both uridine and cytidine could be used as internal standards in vicine and convicine analysis, because they occur mainly as intermediate products in plant metabolism and are therefore not detected in analysis of vicine and convicine. Uridine has similar properties and extractability to vicine and convicine, so it is a good choice as an internal standard.

The formation and stability of the aglycones was studied by conducting the extraction with either 0.05 M sodium phosphate buffer or water, and large molecular weight components, such as proteins and polysaccharides, were removed by centrifuging through regenerated cellulose membrane Amicon® filters. The ȕ-glucosidase was presumably also retained by the filter to stop the hydrolysis. The aglycone extraction was kept as simple as possible to prevent changes induced by pH or temperature. However, vicine, convicine and the aglycones could all be extracted either with buffer or water, as concluded above. The solubilities of divicine and isouramil were similar to the solubilities of vicine and convicine (Marquardt, Frohlich et al. 1989; Marquardt, Arbid et al. 1989); thus, their analysis alongside vicine and convicine is feasible.

6.1.2 Performance of the HPLC analysis for vicine and convicine

The optimised RP-HPLC-UV method was suitable for the analysis of vicine and convicine.

The analytes were separated on a C18 column at 30 °C, and the separation was good with both tested eluents (water with 0.1% formic acid and water alone) at flow rate of 0.8 ml/min.

Vicine and convicine contain amino and hydroxyl groups that affect their properties. Vicine has one more amino group than convicine, so it is more basic than convicine. The pKa1

values for vicine and convicine are 3.16 and 2.71, respectively, according to (ACD/Labs) Software V11.02 (© 1994-2012 ACD/Labs), whereas the pKa value for unsubstituted pyrimidine is 1.3. At an acidic pH, amino groups are protonated. The pH of the eluent containing 0.1% formic acid was 2.7, and the pH of the MilliQ-water eluent was close to neutral. In general, retention on C18 column is better when compounds are in their neutral form. Based on the pKa values, vicine and convicine are partly protonated and neutral in the acidic eluent (pH 2.7). The retention time was shorter for vicine than for convicine with an acidic eluent, but was longer for vicine and slightly shorter for convicine with water as an eluent. Zhang et al (2003) also reported an effect of pH on retention time of vicine. The retention times of the internal standard uridine were less affected by the pH of the eluents.

The pH of the eluent was slightly lower (pH 2) in the previous studies, that used 0.05 M ammonium phosphate buffer (Burbano et al. 1995; Goyoaga et al. 2008; Cardaror-Martinez et al. 2012) as an eluent, following the protocol of Marquardt and Frohlich (1981). Water was used first used as an eluent by Quemener (1988) and then by Griffiths and Ramsay

63

(1992). Overall, the separation was good with both tested eluents. The eluent containing 0.1% formic acid was selected in the present study for the final method, because formic acid (pKa value 3.75) has a buffering capacity to retain retention times, and it enables protonation of analytes for the LC-MS-analysis.

The published absorption maxima for vicine (275 nm), convicine (271 nm) and uridine (261 nm) were used to support the identifications of the studied compounds (Ploeser and Loring 1949; Bendich and Clements 1953; Bien et al. 1968). In addition, MS detection was used to confirm the identity of vicine and convicine based on their known MS spectra. The protonated molecular ions for vicine and convicine, m/z 305 and m/z 306, respectively, were detected in positive ionisation with LC-ESI-MS.

The analytes were quantified by UV detection and showed a wide linear range, as the response was linear for injections of 50 ng up to 7.5 μg. No interfering UV absorbing compounds were detected in the chromatograms. The detection wavelength was selected to be suitable for the internal standard uridine. Quantification using uridine was a practical and precise choice, and standard curves of vicine and convicine were not needed in every analysis set. By contrast, the relative response factors were needed because the response of uridine differed from the responses of vicine and convicine. The relative response factors of uridine to vicine and convicine, as established in study I, were close to each other in acidic media at 273 nm, at 0.65 for vicine and 0.61 for convicine. The earlier established relative response factors vicine and convicine to uridine showed a slightly higher value for vicine (0.8) and a similar value for convicine (0.6) in water at 273 nm when compared to the values found in acidic media in study I (Quemener 1988). Marquardt and Frohlich (1981) stated that the response of convicine would be a bit lower than the response of vicine at concentrations of 52 μg/ml at 280 nm. However, the similarity of the relative response factors of vicine and convicine is reasonable because their molar absorption coefficients are close to each other (Bendich & Clements, 1953; Bien et al, 1968). Based on this finding, both compounds could be quantified using only a vicine or convicine standard, as done in study III.

Overall, the performance of the RP-HPLC-UV analysis method was good, as it was similar or slightly better when compared to earlier methods (Marquardt & Frohlich, 1981;

Quemener 1988; Burbano et al, 1995). The responses were linear across a wide range, and RSD% values presenting the repeatability of the method were 3–7% for vicine and convicine for the in-house reference over a longer time (n=20).

6.1.3 Simultaneous HPLC analysis of the aglycones

The analysis method was first taken in use for quantification of vicine and convicine and later in study II, and its suitability for analysis of the aglycones was confirmed. The aglycones were determined, and further oxidation products were indicated. The main

64

advantage of the method was the capability for simultaneous analysis of vicine, convicine and the aglycone forms.

The aglycones were analysed with the RP-HPLC-UV method, which has not been previously used for that purpose. The identities of divicine and isouramil were confirmed based on their known UV absorption maxima and mass spectra. In the early study by Marquardt and Frohlich (1981), indications of the aglycones were seen when the formation of new peaks was noticed after acid hydrolysis of vicine and convicine. However, the peaks were not identified. An HPLC method with a strong cation exchange system (SCX) was instead applied for analysis of the aglycones, which were studied separately as pure compounds. An SCX-HPLC method was also used to study divicine by Marquardt, Frohlich et al. (1989) and McMillan et al. (1993) and to study isouramil by Marquardt, Arbid et al.

(1989). Further degradation products were also observed with the SCX-HPLC methods.

Separation of the aglycones was slightly better with the SCX-HPLC methods used in the previous studies than with the RP-HPLC method applied in study II, but the SCX-HPLC system was more complicated and would not be compatible with subsequent MS detection for identification. The retention times for divicine were 2–4 min longer with the SCX-HPLC system (Marquardt, Frohlich et al. 1989; McMillan et al. 1993) than in this work with the RP-HPLC system. The oxidised divicine was detected and identified based on the decrease of the divicine peak area after addition of a reducing agent (Marquardt, Frohlich et al. 1989;

McMillan et al. 1993). Oxidised divicine was also determined in study II. Apart from oxidised divicine, one unknown peak was reported by Marquardt, Frohlich et al. (1989) and two unknown peaks by McMillan et al. (1993). In study II, oxidised divicine was the main oxidation product of divicine. In contrast to the findings of Marquardt, Arbid et al. (1989), no further oxidation products were detected from isouramil in the present study. However, the enzymatic hydrolysis used in this work was a milder treatment than the acid hydrolysis used in most of the earlier studies. For example, acid hydrolysis was suggested to induce the formation of deaminated products from divicine (Pedersen et al. 1988; Winterbourn et al. 1989; McMillan et al. 1993). Therefore, the findings cannot be directly compared. In study II, other further products of divicine and isouramil were seen at retention times 1.6-1.9 min, which is not the most optimal for the determination of reactions products because buffers and other slightly retained compounds can interfere with the analysis.

The further reactions of the aglycones could be studied by implementing new methods, such as the HILIC-MS system recently developed for vicine and convicine (Purves et al. 2018a).

An HILIC column was also used for determination of pyrimidines and purines (Marrubini et al. 2010). The retention times for vicine and convicine were short, as they eluted in 3 min with the selected conditions (Purves et al., 2018a). The retention of the aglycones of vicine and convicine on an HILIC column has not yet been studied. MS detection could be used for structure analysis of further reaction products of the aglycones when coupled with suitable chromatographic methods.

65

To conclude, the selected conditions in the analysis method were suitable for the simultaneous study of vicine, convicine, the aglycones and oxidised divicine. This benefit was further utilised in studies II and III to observe the hydrolysis of vicine and convicine and the formation and stability of divicine and isouramil. When studying the aglycones, samples should be immediately analysed or the aglycones should be handled in the absence of oxygen or in the presence of effective reductants.

6.2 Occurrence and variation of vicine and convicine in domestic